From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:39352) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1dcDSj-0004np-BB for qemu-devel@nongnu.org; Mon, 31 Jul 2017 12:22:34 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1dcDSh-0005OA-2o for qemu-devel@nongnu.org; Mon, 31 Jul 2017 12:22:33 -0400 From: Anton Nefedov Date: Mon, 31 Jul 2017 19:21:54 +0300 Message-Id: <1501518125-29851-3-git-send-email-anton.nefedov@virtuozzo.com> In-Reply-To: <1501518125-29851-1-git-send-email-anton.nefedov@virtuozzo.com> References: <1501518125-29851-1-git-send-email-anton.nefedov@virtuozzo.com> MIME-Version: 1.0 Content-Type: text/plain Subject: [Qemu-devel] [PATCH v3 02/13] block: treat BDRV_REQ_ALLOCATE as serialising List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: qemu-block@nongnu.org, den@virtuozzo.com, kwolf@redhat.com, mreitz@redhat.com, eblake@redhat.com, Anton Nefedov The idea is that ALLOCATE requests may overlap with other requests. Reuse the existing block layer infrastructure for serialising requests. Use the following approach: - mark ALLOCATE serialising, so subsequent requests to the area wait - ALLOCATE request itself must never wait if another request is in flight already. Return EAGAIN, let the caller reconsider. Signed-off-by: Anton Nefedov --- block/io.c | 27 +++++++++++++++++++-------- 1 file changed, 19 insertions(+), 8 deletions(-) diff --git a/block/io.c b/block/io.c index b3f9d61..1b935d8 100644 --- a/block/io.c +++ b/block/io.c @@ -512,7 +512,8 @@ void bdrv_dec_in_flight(BlockDriverState *bs) bdrv_wakeup(bs); } -static bool coroutine_fn wait_serialising_requests(BdrvTrackedRequest *self) +static bool coroutine_fn wait_serialising_requests(BdrvTrackedRequest *self, + bool nowait) { BlockDriverState *bs = self->bs; BdrvTrackedRequest *req; @@ -543,11 +544,14 @@ static bool coroutine_fn wait_serialising_requests(BdrvTrackedRequest *self) * will wait for us as soon as it wakes up, then just go on * (instead of producing a deadlock in the former case). */ if (!req->waiting_for) { + waited = true; + if (nowait) { + break; + } self->waiting_for = req; qemu_co_queue_wait(&req->wait_queue, &bs->reqs_lock); self->waiting_for = NULL; retry = true; - waited = true; break; } } @@ -1053,7 +1057,7 @@ static int coroutine_fn bdrv_aligned_preadv(BdrvChild *child, } if (!(flags & BDRV_REQ_NO_SERIALISING)) { - wait_serialising_requests(req); + wait_serialising_requests(req, false); } if (flags & BDRV_REQ_COPY_ON_READ) { @@ -1352,7 +1356,10 @@ static int coroutine_fn bdrv_aligned_pwritev(BdrvChild *child, max_transfer = QEMU_ALIGN_DOWN(MIN_NON_ZERO(bs->bl.max_transfer, INT_MAX), align); - waited = wait_serialising_requests(req); + waited = wait_serialising_requests(req, flags & BDRV_REQ_ALLOCATE); + if (waited && flags & BDRV_REQ_ALLOCATE) { + return -EAGAIN; + } assert(!waited || !req->serialising); assert(req->overlap_offset <= offset); assert(offset + bytes <= req->overlap_offset + req->overlap_bytes); @@ -1456,7 +1463,7 @@ static int coroutine_fn bdrv_co_do_zero_pwritev(BdrvChild *child, /* RMW the unaligned part before head. */ mark_request_serialising(req, align); - wait_serialising_requests(req); + wait_serialising_requests(req, false); bdrv_debug_event(bs, BLKDBG_PWRITEV_RMW_HEAD); ret = bdrv_aligned_preadv(child, req, offset & ~(align - 1), align, align, &local_qiov, 0); @@ -1476,6 +1483,10 @@ static int coroutine_fn bdrv_co_do_zero_pwritev(BdrvChild *child, bytes -= zero_bytes; } + if (flags & BDRV_REQ_ALLOCATE) { + mark_request_serialising(req, align); + } + assert(!bytes || (offset & (align - 1)) == 0); if (bytes >= align) { /* Write the aligned part in the middle. */ @@ -1494,7 +1505,7 @@ static int coroutine_fn bdrv_co_do_zero_pwritev(BdrvChild *child, assert(align == tail_padding_bytes + bytes); /* RMW the unaligned part after tail. */ mark_request_serialising(req, align); - wait_serialising_requests(req); + wait_serialising_requests(req, false); bdrv_debug_event(bs, BLKDBG_PWRITEV_RMW_TAIL); ret = bdrv_aligned_preadv(child, req, offset, align, align, &local_qiov, 0); @@ -1563,7 +1574,7 @@ int coroutine_fn bdrv_co_pwritev(BdrvChild *child, struct iovec head_iov; mark_request_serialising(&req, align); - wait_serialising_requests(&req); + wait_serialising_requests(&req, false); head_buf = qemu_blockalign(bs, align); head_iov = (struct iovec) { @@ -1604,7 +1615,7 @@ int coroutine_fn bdrv_co_pwritev(BdrvChild *child, bool waited; mark_request_serialising(&req, align); - waited = wait_serialising_requests(&req); + waited = wait_serialising_requests(&req, false); assert(!waited || !use_local_qiov); tail_buf = qemu_blockalign(bs, align); -- 2.7.4